I/O Systems Sunu Wibirama If you are obviously IT student Then - - PowerPoint PPT Presentation

I/O Systems

Sunu Wibirama

If you are obviously IT student

Then you should know these pictures...

13.

Introduction



Main job of computer : I/O and processing (the latter is rarely happened)



Browsing: read and enter information, not compute an answer



OS: manage, control I/O devices and operations



What we will explore today:

 I/O Hardware  Application I/O Interface  Kernel I/O Subsystem



What should you read by yourself:

 Transforming I/O Requests to Hardware Operations  STREAMS  Performance

KERNEL

Kernel’s main task: resource management, communication between hardware & software

Kernel is the core of operating system

13.

I/O Hardware



Major concern of OS designer is: I/O support. Why?



Incredible variety of I/O devices: mouse, hard disk, cd-rom, usb, etc.



Common concepts of I/O hardware elements:

 Port / connection point (ex.: serial port, parallel port)  Bus: wires with spesific protocol  Controller (host adapter) : collection of electronics that can

• perate a port, a bus, or a device.

Serial-port controller v.s. SCSI controller



Details are encapsulated in device-driver module in OS kernel



Device drivers: uniform I/O instructions to control I/O devices



Controller has one or more registers for data and control signals



How processor give commands / data to a controller for I/O transfer?

 Direct I/O instructions using bus: send special instructions to an I/O port

• address. Must select proper device and send data to device register

 Memory-mapped I/O: device registers are mapped into the address

space of the processor

A Typical PC Bus Structure

PCI bus - connected to fast devices Expansion bus - connected to slow devices

13.

Device I/O Port Locations on PCs (partial)

Graphic controller : memory-mapped I/O to hold screen contents, because writing data into memory-mapped region is faster than issuing millions of instructions

13.

I/O Port’s Register



Data-in register: read by the host to get input



Data-out register: written by the host to send output



Status register: contain status bits that can be read by the host



Control register: can be written by the host to start command or to change the mode of the device. Example:

 Full-duplex or half-duplex mode;

Duplex: communicate (send and receive message). Half-duplex: only one activity at one time. Full-duplex: can either send or receive information at the same time.

13.

I/O Transfer Control

 Determine when the next data transfer occurs:

 Polling  Interrupts

 Direct Memory Access (DMA)

13.

Polling



CPU (Host) keeps checking I/O Controller status register until detect new input data arrives OR output devices ready for next data.



Going on a trip an ask: “Are we there yet?” “Are we there yet?” “Are we there yet?”



Controller:

 Status register: 1 - busy and 0 - clear



Host:

 Command register: command-ready bit



Handshaking step (repeated for each byte):

 Host repeatedly read busy bit until that bit becomes clear (busy waiting or polling)  Host sets the write bit in the command register and writes a byte into the data-out

register

 Host sets command-ready bit  Controller notices that command-ready bit is set, it sets the busy bit  Controller reads the command register, sees the write command. It reads data-out

register to get the byte and does the I/O to the device

 Controller clears command-ready bit, clears error bit in the status register to

indicate I/O succeeded, clears the busy bit to indicate that I/O operation is finished

13.

Interrupt

 Polling is inefficient if rarely find idle devices, but large

tasks remain undone. How to solve?

 How about: device sends special signal to CPU when

new data arrives OR device ready for next data

 Going on a trip an ask: “Wake me when we get there!”  Interrupt : Notify CPU if device is ready, instead of

polling repeatedly

13.

Interrupt-Driven I/O Cycle

13.

Direct Memory Access (DMA)



Large data transfer: disk drive



CPU is kept from exhausted work: avoid programmed I/O for large data movement (programmed I/O = watch status bits and to feed data into a controller register one byte at a time)



How? Bypasses CPU to transfer data directly between I/O device and

• memory. Delegating CPU tasks to DMA controller.



Initiate DMA transfer:

 CPU writes a DMA command block into memory  This command block contains: pointer to the source of transfer,

pointer to the destination of transfer, and a count of the number of bytes to be transferred

 CPU writes the address of command block to DMA controller and

let the DMA controller do the rest

13.

Six Step Process to Perform DMA Transfer

13.

Application I/O Interface



Explain interface and techniques of OS to treat I/O devices in standard way (such as open a file on a disk without knowing what kind of disk it is).



I/O system calls encapsulate device behaviors in generic classes



Device-driver layer hides differences among I/O controllers from kernel



Devices vary in many dimensions

 Character-stream or block  Sequential or random-access  Sharable or dedicated  Speed of operation  read-write, read only, or write only

13.

Characteristics of I/O Devices

13.

A Kernel I/O Structure

uniformity

think the structure as organization structure: head of organization doesn’t need to know the detail

13.

Kernel I/O Subsystem



Kernel provides many services to I/O, such as:

 scheduling  buffering  caching  spooling  device reservation  error handling



These are provided by I/O subsystem, build on hardware and device-driver infrastructure

13.

Kernel I/O Subsystem



Scheduling

 Meaning: to determine a good order to execute I/O requests  Can improve overall system performance  Some I/O request ordering via per-device queue  Some OSs try fairness



Buffering - store data in memory while transferring between devices

 To deal with device speed mismatch (modem v.s. storage)  To deal with device transfer size mismatch (large data re-assembling

in computer networking)

 To maintain “copy semantics” (data reliability when write() system call is

• executed. Copying an opened large Ms. Word file to USB flash

drive)

13.

Sun Enterprise 6000 Device-Transfer Rates

Logarithmic scale

13.

Kernel I/O Subsystem



Caching - fast memory holding copy of data (do you remember the function of proxy

server in networking? - you need to press CTRL+F5 to renew the cache)  Always just a copy (not the only existing item - as buffer does)  Key to performance



Spooling - hold output for a device

 If device can serve only one request at a time  i.e., Printing - no interleave allowed



Device reservation - provides exclusive access to a device

 System calls for allocation and deallocation  Watch out for deadlock

13.

Error Handling



OS can recover from disk read, device unavailable or defective, transient write failures



Most return an error number or code when I/O request fails

 Unix : errno 2 - no such file or directory



System error logs hold problem reports

13.

I/O Protection



User process may accidentally or purposefully attempt to disrupt normal operation via illegal I/O instructions

 All I/O instructions defined to

be privileged: I/O must be performed via OSʼ system calls

 Memory-mapped and I/O

port memory locations must be protected too (ex.: graphic games and video editing should access the memory

• directly. OS should protects a

block of memory for graphic processing at one time)

Memory layout

13.

Kernel Data Structures



Kernel keeps state info for I/O components, including open file tables, network connections, character device state



Many, many complex data structures to track buffers, memory allocation, “dirty” blocks



Unix uses object-oriented methods to encapsulate “differences” (for example: read() operation, the semantics is different for each I/O devices)



Windows uses message passing to implement I/O

 I/O request is converted into a message through the kernel to the

I/O manager and then to device driver, each of which may change the message contents.

 For output, the message contains data to be written. For input,

the message contains buffer to receive data.

Life cycle

• f I/O Request

(using DMA)

Thank You